Low latency applications are highly sensitive to the overhead introduced by virtualization and container orchestration. This paper evaluates the combined effects of Kubernetes and DPDK on a Layer 2 packet reflector, implemented in both Python and DPDK.Across six deployment scenarios - including bare-metal and containerized setups with and without Multus CNI - we measure the impact on latency and jitter. Results show that Kubernetes introduces significant overhead, while DPDK reduces latency by up to 70% and jitter by 55%, restoring near-native performance even in containerized environments.Our findings demonstrate that DPDK, when paired with direct NIC access through Multus, is an effective solution for deploying real-time services in modern cloud-native infrastructures.
Low Latency in Containerized Environments: A Performance Analysis with DPDK and Kubernetes
Ramadan, Ibrahim Mohamed
;Centofanti, Carlo;Marotta, Andrea;Graziosi, Fabio
2025-01-01
Abstract
Low latency applications are highly sensitive to the overhead introduced by virtualization and container orchestration. This paper evaluates the combined effects of Kubernetes and DPDK on a Layer 2 packet reflector, implemented in both Python and DPDK.Across six deployment scenarios - including bare-metal and containerized setups with and without Multus CNI - we measure the impact on latency and jitter. Results show that Kubernetes introduces significant overhead, while DPDK reduces latency by up to 70% and jitter by 55%, restoring near-native performance even in containerized environments.Our findings demonstrate that DPDK, when paired with direct NIC access through Multus, is an effective solution for deploying real-time services in modern cloud-native infrastructures.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


